AerialMegaDepth: Aerial and Ground Images for 3D Reconstruction

AerialMegaDepth is a dataset for learning aerial-ground scene reconstruction and novel view synthesis, presented at CVPR 2025. The dataset was created by Khiem Vuong, Anurag Ghosh, Deva Ramanan, Srinivasa Narasimhan, and Shubham Tulsiani. It was last updated on the Hugging Face platform on 2025-09-30.

Use Cases

• Training 3D reconstruction models based on paired aerial and ground imagery.
• Developing novel view synthesis algorithms for aerial scenes.
• Researching multi-view geometry and depth estimation from different perspectives.

Strengths

• Dataset is associated with a CVPR 2025 publication, indicating peer-reviewed research context.
• Last updated on 2025-09-30, suggesting recent maintenance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and dataset size are unknown, which may limit suitability assessment.

Provenance

Source

huggingface

Freshness

Last updated 2025-09-30 21:09:11